Spectral Decomposition of Surface Ozone Variability

The concentrations of atmospheric species vary on hourly to decadal timescales, driven by a range of chemical and physical processes. Using spectral analysis it is possible to decompose a time series into a set of cycles with differing frequency, amplitude and phase. Traditional Fourier analysis is inappropriate as it is susceptible to artifacts due to gaps in typical atmospheric observational datasets; instead we use the Lomb-Scargle Periodogram, allowing spectral analysis of gapped data. We exploit this method to construct metrics for the evaluation of atmospheric chemistry models.

Spectral analysis of global surface ozone observations from networks of ground-based instruments (GAW (global), EMEP (Europe), CASTNET (USA) etc.) reveal characteristic patterns of influence on ozone, at turbulent, synoptic and macro-weather timescales, and significant periodicity at the diurnal, semi-annual and annual timescales. We sample archived hourly surface ozone fields from a chemistry transport model (GEOS-Chem; driven by observed weather) and a chemistry-climate model (GFDL CM3; generates its own weather) at the location of the measurements, and evaluate them with the observation-derived power spectra. Unsurprisingly, given their temporal and spatial resolution, the models underestimate power on the turbulent timescale but demonstrate skill on the synoptic and macro-weather scales. The distinct diurnal, semi-annual and annual periodicities in the observations are evident in the models also, but with varying degrees of accuracy.

The use of spectral analysis in model analysis forms a novel method for the systematic and quantitative evaluation of Chemical Transport / Earth System Models on a range of timescales, and may prove useful for evaluating multi-model simulations coordinated under the Chemistry Climate Modeling Initiative.